Predicting Treatment Outcomes, Attrition, and Enrolment Suitability in Mindfulness-integrated Cognitive Behavior Therapy: Machine Learning and Survival Analyses
Sarah Strohmaier, Bruno Cayoun, Alice Shires, Christian U. Krägeloh, Oleg N. Medvedev
July 18, 2026 preprint DOI: 10.31234/osf.io/gzahr_v1 (opens in new tab)
Study at a glance
AI-extracted from the abstract| Characteristics | Observational cohort |
|---|---|
| Sample size | 326 |
| Population | Adults assessed at two psychology clinics |
| Intervention | Mindfulness-integrated Cognitive Behavior Therapy |
| Topics | Meditation |
| Keywords | Distress Life satisfaction Logistic regression Depression economics Cognition Personality Discontinuation Cognitive therapy Cognitive flexibility Cognitive behavioral therapy Clinical psychology Medical diagnosis Baseline sea Multiple baseline design |
| Key findings | MiCBT produced large treatment effects on stress and equanimity, with depression severity, PTSD, ADHD, age, life satisfaction, and personality disorder diagnosis predicting different aspects of enrolment, improvement, and retention. |
Abstract
Objectives: Mindfulness-integrated Cognitive Behavior Therapy (MiCBT) is an evidence-based, structured, transdiagnostic program. This study investigated three interrelated questions: (a) which characteristics predict initial enrolment suitability following intake assessment. (b) which baseline characteristics predict clinical improvement during MiCBT, (c) which characteristics predict treatment retention.Method: Participants were 326 adults (195 male; Mage = 41.4, SD = 14.0) assessed at two psychology clinics. Of 326 assessed at intake (baseline), 190 engaged in MiCBT (pre-treatment). Mixed-effects growth models, Elastic Net, logistic regression, and XGBoost estimated enrolment and treatment outcomes, while Cox regression, Random Survival Forests, and Accelerated Failure Time estimated retention.Results: Large treatment effects were observed (Stress d = 1.47, Equanimity d = 1.71). More severe Depression and PTSD diagnoses moderated faster distress reduction, while PTSD and ADHD predicted greater life satisfaction gains. Equanimity improvement was strongly correlated with distress reduction (r = –0.47) and life satisfaction gains (r = 0.39). For retention, younger age (HR = 0.81, p = 0.027), lower life satisfaction at baseline (HR = 0.83, p = 0.041), and personality disorder diagnosis (HR = 1.84, p = 0.042) were significant predictors. Critically, prior meditation experience predicted higher likelihood of enrolment (χ² = 9.44, p = 0.002) but not subsequent retention (p = 0.885), while diagnosis of ADHD predicted non-enrolment (χ² = 6.69, p = 0.010) but not discontinuation among those who had commenced.